/reflect

Audit recent workspace context and suggest memory updates, skills, commands, and workflows.

What This Command Does

/reflect reviews recent accessible workspace context and returns a reflection report. It looks for stale assumptions, missing durable memory, repeated agent mistakes, duplicated work, and opportunities to create reusable skills, Prompt Commands, or workflows.

This is a shipped Dawn Prompt Command. It starts a normal agent run with stored instructions attached, so the selected agent still controls available tools, memory, skills, connected sources, and permissions.

Syntax

/reflect <optional focus, timeframe, team, workflow, artifact, or question>

Arguments

Use the text after /reflect as an optional focus. If you leave it blank, Dawn inspects recent high-signal workspace context.

Examples:

/reflect onboarding and setup friction
/reflect check memory for stale Azure DevOps assumptions
/reflect after the billing migration

Execution

  • Default mode: background
  • Preferred provider: OpenAI
  • Preferred model: gpt-5.4-mini
  • Tags: workspace, memory, audit, workflow

Background commands create background work where the surface supports it. Foreground commands run in the active conversation path.

What The Report Includes

The command returns a reflection report with sections for summary, findings, suggested memory updates, suggested skills, suggested commands, suggested workflows, open questions, and next actions.

Substantive findings should include evidence such as a conversation title or ID, message timestamp, memory page title, skill or command slug, workflow name, or an explicit note when a claim is only a hypothesis.

Read-Only By Default

/reflect is proposal-only. It can read accessible conversations, memory pages, skills, commands, agents, and workflows, but it should not update memory, create skills, edit commands, start workflows, or post to connected sources during the audit run.

Review the report before applying any suggested memory, skill, command, or workflow changes.

Tool Guidance

  • search_conversation_history
  • get_conversation_thread
  • list_memory_pages
  • search_memory_pages
  • read_memory_page
  • search_my_memory
  • read_my_memory_page
  • search_workspace_skills
  • read_workspace_skill
  • validate_workspace_skill_draft
  • search_workspace_commands
  • read_workspace_command
  • validate_workspace_command_draft
  • list_agents
  • read_agent_capabilities
  • list_workflows
  • get_workflow

Tools listed here are command guidance. They do not bypass the selected agent’s source assignments, policy, or workspace permissions.